Limited water quality data is often responsible for incorrect model description and misleading interpretation in water resources planning and management scenarios. This study compares two hybrid strategies to convert discrete concentration data into continuous daily values for one year in different river sections. Model A is based on an autoregressive process, accounting for serial correlation, water quality historical characteristics (mean and standard deviation) and random variability; the second approach (model B) is a regression model, based on the relationship between monitoring flow and concentrations, plus an error term. The generated series (here referred to as synthetic series) are propagated in time and space by a full deterministic model (SihQual), that solves the Saint-Venant and advection-dispersion-reaction equations. Results reveal that both approaches are appropriate to reproduce the variability of biochemical oxygen demand and organic nitrogen concentrations, leading to the conclusion that the combination of deterministic/empirical and stochastic components are compatible. A second outcome arises from the comparison of results in different time scales, supporting the need for further assessment of statistical characteristics of water quality data - which relies on monitoring plans. Nonetheless, the proposed methods are suitable to estimate multiple scenarios of interest in water resources planning and management.